Skip to main content ↑ ↓

Notice

Please note that most of the software linked on this forum is likely to be safe to use. If you are unsure, feel free to ask in the relevant topics, or send a private message to an administrator or moderator. To help curb the problems of false positives, or in the event that you do find actual malware, you can contribute through the article linked here.
Topic: Vinyl vs Digital and 24 bit vs 16 bit from vinyl. (Read 206251 times) previous topic - next topic
0 Members and 1 Guest are viewing this topic.

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #75 –
Quote from: Axon on
... That said, I do intend on scoring a 78 cart someday, and unless I want to spend $2000 for an EMT, or get a retip on a DL-102 for some smaller but still absurd price, I'll probably get an Ortofon or Shure 78 cartridge, and wire up an inline capacitive network with XLR connectors.

I don't want to poke my nose in if you've got a plan for transcribing 78's but I've had quite a lot of experience with them - I've restored getting on for 200, with varying degrees of success. So, if you think it might help I'll gladly share my experience with you. I'm not sure such a discussion fits with this thread so maybe you could start a new one or just PM me

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #76 –
Sure, thanks - although I'll warn you, this is a very long term goal of mine. I don't own any 78s, alas.

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #77 –
Quote from: Slipstreem on
Quote from: Arnold B. Krueger on
Never had any problem with applying loading capacitance at the preamp end. Indeed, my main phono preamp in the days of vinyl was based on NE5534 ICs, and had tuning capacitors wired across the inputs.



All coaxial cable inherently has inductive properties as well as capacitive,


However, the inductance of a cable can be quantified.  The capacitance  of a shielded cable will be from 15 to 60 pF per foot, and the inductance will be on the order of  2 tenths of a microhenry per foot.  In contrast the inductance of a MM cartrdige is about half a henry (IOW 500 millihenries).  The difference is a ratio of 250,000, which is to say that the inductance of a few feet of coax cable is negligable compared to that of the other important inductance in the system.

Quote
so placing the capacitors at the far end of a cable run is effectively placing another L/C circuit in series/parallel with the cartridge before the loading capacitors.


However this additional LC circuit resonates so far out of the audio band, and has such low Q, that it is totally negligable. If it mattered there would be some evidence in the frequency response curve of the cartridge, and there never is.

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #78 –
Quote from: Axon on
Quote from: Arnold B. Krueger on
So, educate me! ;-)

Dynamic range has two parts - noise floor and headroom. YOu seem to be OK on the noise floor end, but what's happening with headroom?  Before you apply eq in the digital domain, how close do peak levels come to FS?


Like I said, it clips on rare occasion - although usually this happens either with a pop/tick or when I'm playing a 33 at 45rpm by accident, or when significant groove wear is already evident in the final product - cases that I largely don't care about.


Well, that is evidence of the potential dynamic range problem I was talking about. Traditional feedback RIAA circuits put the noise floor at least 10 dB below quiet groove noise, and put preamp clipping at least 10 dB above the absolute worst case input that can ever possibly happen.

Quote
Otherwise, the clip is for only one or two samples out of every 50 or so. So I'm not too worried about it.


Normally the standard for professional recording is no clip, no place, ever, and about 10-20 dB of headroom on top of that.

Quote
Moreover, this is at the maximum gain setting for the preamp (50db), and I can always turn it down if it becomes an issue.


I'd turn it down until the noise floor becomes a problem. A little noise is an inconvenience, clipping is a disaster.

Quote
And note, this is for a LOMC. Gain goes a lot farther in the flat domain than in the RIAA domain.


It should be all the same. The problem with a flat preamp is overload at mid and high frequencies. If I were designing a playback system for LPs that was centerpieced by digital equalization, I'd consider building it with a - 6 dB/octave roll-off starting at 50 Hz, and putting in the midrange turnover in with digital equalization.

The way I'd set it up is play back a well-respected test record and simply set the default eq so that the FR bands on the test record play back as specified.

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #79 –
Quote from: Arnold B. Krueger on
However, the inductance of a cable can be quantified.  The capacitance  of a shielded cable will be from 15 to 60 pF per foot, and the inductance will be on the order of  2 tenths of a microhenry per foot.  In contrast the inductance of a MM cartrdige is about half a henry (IOW 500 millihenries).  The difference is a ratio of 250,000, which is to say that the inductance of a few feet of coax cable is negligable compared to that of the other important inductance in the system.
From memory, I believe that the inductance of the cartridge in question was closer to 50mH than 500, but this still leaves only a miniscule effect within the audio band that certainly wouldn't have been audible. It seems as though I was suckered by our old friend Mr Placebo, so I retract my previous claim of there being an audible difference.

My notes on the design, building and testing of the scratch eliminator in question are 25 years old and I now believe that my hand-drawn sketches in the brief notes relating to ringing were taken after the channel differential of an 18dB/octave high-pass filtering at 20kHz of the non-RIAA corrected signal was undertaken. No low-pass filtering was applied above 20kHz, so the resultant sketch contained everything from 20kHz upwards and hardly anything below it.

The effects outside the normal audio pass-band were significant enough to make scratch detection considerably easier with my particular cartridge and detector arrangement, but the effect of capacitive loading placement obviously doesn't bear any relevance under normal listening conditions as was my previous mistaken assumption.

Thanks for clarifying this for me.

Cheers, Slipstreem. 

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #80 –
Quote from: Arnold B. Krueger on
Well, that is evidence of the potential dynamic range problem I was talking about. Traditional feedback RIAA circuits put the noise floor at least 10 dB below quiet groove noise, and put preamp clipping at least 10 dB above the absolute worst case input that can ever possibly happen.
I disagree. The "native" noise floor of the ADC is in the -100db range; the -72db noise that exists is almost entirely due to the preamp. That clipping simply reflects my laziness in not turning the gain down to reduce the risk of clipping. That said, that points out one sorta significant disadvantage to using mic preamps: their gains can't be ganged together, so every time I want to change the gain, I need to schlep to my test record and make sure the knobs are set correctly.

Quote
Normally the standard for professional recording is no clip, no place, ever, and about 10-20 dB of headroom on top of that.
Yes, but I'm aiming for quality and laziness, not professionalism.  I am utterly confident that I would not be able to pass an ABX test if a single-sample peak, at 96khz, was clipped by 10% or 20% in such a fashion.

Quote
It should be all the same. The problem with a flat preamp is overload at mid and high frequencies. If I were designing a playback system for LPs that was centerpieced by digital equalization, I'd consider building it with a - 6 dB/octave roll-off starting at 50 Hz, and putting in the midrange turnover in with digital equalization.  The way I'd set it up is play back a well-respected test record and simply set the default eq so that the FR bands on the test record play back as specified.
Rob over at Pure Vinyl actually does have a software setting for if one's mic preamp has single pole or zero of the RIAA eq wired into it - the 75us one, as I recall. It's largely for the same reasons.

I'm not disputing anything you say on this, but I'll admit that my reasons for going flat are largely intrinsic rather than for particularly sound reasons - simply put, I wanted a test platform to do recordings with as clean a transient response as possible, and eliminating all equalization entirely from the circuit is advantageous in that situation.


Also, if I may ask: What is your opinion of the OC9?

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #82 –
Quote from: Arnold B. Krueger on
Quote from: Axon on

Also, if I may ask: What is your opinion of the OC9?


No experience, no opinion.
Really?  (Sorry - didn't mention my alter egos on other sites, and I forgot that you mentioned that on Usenet as opposed to here. Kind of a brain fart that I even asked that here.)

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #83 –
Goodness, Axon on RAHE, Arny on HA, a mixed-race Democrat elected President -- it's a world gone mad!

I like it. 

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #84 –
I feel old.  Good to see you Arny.

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #86 –
Quote from: krabapple on
Goodness, Axon on RAHE, Arny on HA, a mixed-race Democrat elected President -- it's a world gone mad!

I like it. 



Last I looked RAHE was dead, the only remaining moderator wanted me to apologize to him for pointing out tht he was hideously biased in favor of terminal woo, and a couple of the notables there were holding forth that I had never run a DBT in my life, and that I didn't know what an ABX test was.

Vinyl, still, has rather some "good sounding" distortions, for some people, at least, but they are distortions.

Vinyl also appears, when you mistrack, to have a huge dynamic range. This is solely due to the fact that mistracking creates huge outputs from the vinyl, and does not reflect on its signal carrying capacity.

As to digital capture of vinyl, if one wants to see what kind of SNR is lost in the PCM range when capturing without RIAA, one simply has to look at the limits of the RIAA curve, take the difference, and that's how many dB you lose. Does it compare to a phono-preamp? Depends on the phono preamp. Mine, at least, does a better job than capturing vinyl at 20 bits and then doing RIAA EQ.
-----
J. D. (jj) Johnston

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #87 –
Wow, talk about bum luck on RAHE. Or just bums.

I tend to agree about dyanamic range and vinyl, although I would point out that the false dynamic range is probably due to mistracing (including due to wear) and not mistracking. At least in mainstream listening situations (records cut near 0db, 9" arms, reasonable alignment, no skipping).

Quote
As to digital capture of vinyl, if one wants to see what kind of SNR is lost in the PCM range when capturing without RIAA, one simply has to look at the limits of the RIAA curve, take the difference, and that's how many dB you lose. Does it compare to a phono-preamp? Depends on the phono preamp. Mine, at least, does a better job than capturing vinyl at 20 bits and then doing RIAA EQ.
But what's the peak spot signal power? Is it even physically possible for a record to hit 0db on your preamp with a 20khz sine?

At my current gain levels on my flat setup, 0db @1khz hits at around -33dbFS. So the "weakest" signal that would cause clipping would be around 20khz +13db. I assert that is untrackable. In that context, SNR needs to be evaluated against both a variable signal ceiling (relating to maximum tracking ability, and also to the maximum levels that are typically encountered in cut records) and a variable noise floor. Given the acceleration and velocity constraints of the medium and the transducers it is entirely reasonable to believe that the signal limits become more constrained as frequency increases.

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #88 –
Quote from: Woodinville on
Quote from: krabapple on

Goodness, Axon on RAHE, Arny on HA, a mixed-race Democrat elected President -- it's a world gone mad!

I like it. 



Last I looked RAHE was dead, the only remaining moderator wanted me to apologize to him for pointing out tht he was hideously biased in favor of terminal woo, and a couple of the notables there were holding forth that I had never run a DBT in my life, and that I didn't know what an ABX test was.


Must've been some time ago....there's now two moderators, though the previous one never struck me as biased in favor of woo...if anything, for the last few years he pretty much let me and a few others rail against it at whim, as long as we actually make an argument, and don't get personal.  Today it's not quite dead, though perhaps it smells funny...the posting roster is much reduced to maybe a dozen or so 'notables' from its heyday of years back, and I wouldn't claim any new ground ever really gets laid down.  The liveliest thread right now involves whether audible distortion is INTRINSIC to vinyl production and playback.  The competing answers are

yes it's unavoidable

yes but it's euphonic so that's ok

no it's not intrinsic (the latter accompanied by 'show me exact figure in a JAES paper that proves me wrong').

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #89 –
Quote from: krabapple on
Today it's not quite dead, though perhaps it smells funny...the posting roster is much reduced to maybe a dozen or so 'notables' from its heyday of years back, and I wouldn't claim any new ground ever really gets laid down.


The problem being that many of the posters are very poorly informed, and have heads that seem to be made out of tungsten.

Quote
The liveliest thread right now involves whether audible distortion is INTRINSIC to vinyl production and playback.  The competing answers are

yes it's unavoidable  [that has to be what you and I are flogging]

yes but it's euphonic so that's ok

no it's not intrinsic (the latter accompanied by 'show me exact figure in a JAES paper that proves me wrong').

So we cited the papers and then got whined at for not quoting line, chapter and verse.

BTW, you forgot

"Digital has to sound bad because of the empty spaces between the quantization levels causes audible truncation of fading sounds"

Also we have a thread about how changing the power supply on a computer audio interface

"Sound was much smoother, especially in the high frequencies.
The EMU seemed to be able to resolve more as well.
Also, the quiet passages were, um, quieter:)"

I asked the guy to take 5 minutes to run a Rightmark, and of course he said:

"I suppose I could do that, but I do not need to. I know there was a
difference and I'm not about to clarify my results by posting the
technical papers in some science journal."

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #90 –
Quote from: Axon on
I tend to agree about dyanamic range and vinyl, although I would point out that the false dynamic range is probably due to mistracing (including due to wear) and not mistracking. At least in mainstream listening situations (records cut near 0db, 9" arms, reasonable alignment, no skipping).


I think that the perception of dynamic range may be about more than just increasing nonlinear distortion creating a false impression of loudness.

I think that when people hear fades on vinyl, they tend to hear fades into the relatively high mechanically-generated noise floor  that is part of the LP playback system and therefore has a different spatial focussing and spectral content than acoustical noise, which is what you hear with a CD.  With a CD, natural acoustical music fades are going to finish in either the acoustical noise floor of the listening room or the acoustical noise floor of the recording space.

Fades on a well-made CD can be a different experience from the LP due to the mediums inherently lower noise floor. The media-genearted noise floor is so much louder on a LP than on a CD that fades on the CD end up finishing in either the noise floor of the recording space or the listening room space. In either case the noise floor has a different and spectrum than electronic or mechanical noise.

Audiophiles, whether they believe it or not have been subjected to a lot of false propaganda about digital.

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #91 –
Quote from: Arnold B. Krueger on
I think that the perception of dynamic range may be about more than just increasing nonlinear distortion creating a false impression of loudness.

You know, strictly speaking, increasing nonlinear distortion creates a true impression of increased loudness. The loudness actually does go up. It is merely illusory in the sense that it doesn't exist in the original music.

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #92 –
Quote from: Axon on
Quote from: Arnold B. Krueger on
I think that the perception of dynamic range may be about more than just increasing nonlinear distortion creating a false impression of loudness.

You know, strictly speaking, increasing nonlinear distortion creates a true impression of increased loudness. The loudness actually does go up. It is merely illusory in the sense that it doesn't exist in the original music.


aka...euphonic distortion!

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #93 –
Quote from: Arnold B. Krueger on
I think that the perception of dynamic range may be about more than just increasing nonlinear distortion creating a false impression of loudness.


Well, if you don't mind, Arny, perhaps we could discuss the actual effects of distortion that increases with level. 

It does, factually, increase loudness when it spreads the spectrum or creates new spectral elements. What the listener hears is LOUDER, by the real definition of LOUDNESS.

"false" is an interesting term, if you mean "it's not in the original", well, yeah. It's not. But it's in the playback. Could the producer use this? Yes, I think so, although perhaps not knowingly.

If you mean "it's false to the listener", no, spreading the spectrum increases loudness if done in the way that LP's typically distort.

In other words, it's euphonic distortion THAT SOME PEOPLE PREFER. 

That doesn't make it right or wrong. Btw, it's not that hard nowadays to just mimic that distortion digitally.

Btw, the fading into noise instead of digital black also has some the same effects, as does the out-of-phase rumble component from an LP, the interchannel intermodulation, and all those other things that "weak-minded" people who "broke and ran" pointed out years ago.
-----
J. D. (jj) Johnston

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #94 –
[quote name='krabapple' date='Nov 13 2008, 22:06' post='599166']
[quote name='Axon' post='599105' date='Nov 13 2008, 14:42']
[quote name='Arnold B. Krueger' post='599031' date='Nov 13 2008, 07:17']I think that the perception of dynamic range may be about more than just increasing nonlinear distortion creating a false impression of loudness.[/quote]
You know, strictly speaking, increasing nonlinear distortion creates a true impression of increased loudness. The loudness actually does go up.
[/quote]

When I tried to ABX this, I found no way to add distortion and obtain the same sound as a louder, undistorted sound.

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #95 –
Quote from: Arnold B. Krueger on
When I tried to ABX this, I found no way to add distortion and obtain the same sound as a louder, undistorted sound.


Obviously, if something is "louder" the psychological equivalent of "level matching" fails, doesn't it?

Now, of course, you'd need to detail your particular distortion mechanisms, etc, as well.

Would you like to?
-----
J. D. (jj) Johnston

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #96 –
Quote from: Woodinville on
Quote from: Arnold B. Krueger on
When I tried to ABX this, I found no way to add distortion and obtain the same sound as a louder, undistorted sound.


Obviously, if something is "louder" the psychological equivalent of "level matching" fails, doesn't it?


I fear we are not communicating at all.

Quote
Now, of course, you'd need to detail your particular distortion mechanisms, etc, as well.


Method.

Take a typical musical selection a few seconds long. Call this "B" (for Beautiful)

Copy it.

ABX it to see if the copies match.

If they do, then attenuate the copy by 3 dB and ABX the pair again to see if you can tell the difference that the level mismatch makes.

If you can tell the difference reliably, take the attenuated copy  add various amounts of second harmonic distortion. Call these your trial "A"samples (for Awful).

Determine how much distortion it takes to make the anyh of the "A"s  sound the same as  "B".

If you add a little distortion, then it doesn't bring up the level of the "A" sample, and if you add a lot then the "A" sample sounds like it is harsher.

Conclusion: You may make something sound harsher, and some listeners may equate harsher with louder, but in the end the educated ear is not fooled.

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #97 –
Quote from: Arnold B. Krueger on
Quote from: Woodinville on

Quote from: Arnold B. Krueger on
When I tried to ABX this, I found no way to add distortion and obtain the same sound as a louder, undistorted sound.


Obviously, if something is "louder" the psychological equivalent of "level matching" fails, doesn't it?


I fear we are not communicating at all.


That's become obvious. Given that most any level of distortion will create more loudness at some level, I fail to see how you do a "level-matched" test. What DO you define as level-matched, then, pray tell?
Quote
Quote

Now, of course, you'd need to detail your particular distortion mechanisms, etc, as well.


Method.

Take a typical musical selection a few seconds long. Call this "B" (for Beautiful)

Copy it.

ABX it to see if the copies match.

If they do, then attenuate the copy by 3 dB and ABX the pair again to see if you can tell the difference that the level mismatch makes.

If you can tell the difference reliably, take the attenuated copy  add various amounts of second harmonic distortion. Call these your trial "A"samples (for Awful).


So, then, this is done in analog or digital domains? If you did it in the digital domain, you've made an awful mistake and created a mass of aliased inharmonic components that are going to sound "very odd" at the least, unless you've done more than you indicated here. Did you oversample? Did you control your nonlinearities so that no products of the nonlinearity exceed fs/2?

So which did you do? What would the polynomial representation of your distortion mechanism look like?  What was the sampling rate multiplier?

Did you split it into M/S and use different distortions on M/S as is appropriate?
Quote
Determine how much distortion it takes to make the anyh of the "A"s  sound the same as  "B".

How is this remotely germane to this discussion?  You're trying to compare short-term loudness differences on peaks (dynamic range perception) to long-term loudness perception?  You've completely confused the entire issue, and confounded your test with something that just frankly seems, well, irrelevant.
Quote
If you add a little distortion, then it doesn't bring up the level of the "A" sample, and if you add a lot then the "A" sample sounds like it is harsher.

Since, then, we're talking about apparent dynamic range, and  in particular instantaneous loudness, you're not even testing the known facts. I have no idea, what, really, you are testing, but I have little doubt that if you raise the AVERAGE loudness by a factor of 1.21, give or take, by distortion mechanisms, it's going to sound bizzare.

Of course, this has nothing whatsoever to do with the actual effects of loudness enhancement in a dynamic signal. So what have we proven here?
Quote
Conclusion: You may make something sound harsher, and some listeners may equate harsher with louder, but in the end the educated ear is not fooled.


No, you have to provide affirmative evidence for your assertion, not negative evidence for something else. So you have no reasonable justification for that issue.

Now, you have to figure out how to:

Compare perceived dynamic range of two signals via blind test of low-order distortions.
Find out how much dynamic range enhancement sounds "good" and how much sounds "bad".
Figure out how to avoid aliasing in your distortion mechanisms.
Figure out what distortions to use in M and S if you're trying to mimic vinyl.
Figure out what level of antiphase low-frequency distortion is relevant.
Figure out what level of low-level noise, and of what frequency, to use, if LP is what you're mimicing.

The job is not as simple as you appear to be claiming.
-----
J. D. (jj) Johnston

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #98 –
Quote from: Woodinville on
So, then, this is done in analog or digital domains?

Doesn't matter but I usually do stuff like this in the digital domain because its easier.


If you did it in the digital domain, you've made an awful mistake and created a mass of aliased inharmonic components that are going to sound "very odd" at the least, unless you've done more than you indicated here. Did you oversample? Did you control your nonlinearities so that no products of the nonlinearity exceed fs/2?

I don't know who you think you are talking to, but you are obviously talking down/talking trash. But, I'll humor you with some straight answers to show good faith.

(q) Did I oversample?

(a) Don't have to. I just used a sufficiently high sample rate that the signfiicant harmonics were in the usual operational range of the sample rate I used.

I know that there are people who scare small boys by pointing out that out-of-band harmonics that are generated in the digital domain are aliased.  If you are generating highly distorted waveforms with program material that has tons of high frequency energy using say 44/16 then there can be audible artifacts. If you are applying a few percent distortion to regular music, the artifacts are usually not significant.

I typically do work like this at 192/24 or higher, and downsample to 44/16 for delivery. I also carefully check my work with test signals to make sure that practical = theoretical. I've been known to use sample rates up to 10 MHz if my testing deems it important.


So which did you do? What would the polynomial representation of your distortion mechanism look like?

In this case I was working with added second hamonic distortion (an ideal SET) or sharp clipping (an amp that is clean within its ratings, but under rated).


What was the sample rate multiplier?

Since I know there is nothing magic about integer multiplers for upsampling, I used whatever it takes to get to 192/24.


Did you split it into M/S and use different distortions on M/S as is appropriate?

That isn't what happens in the real world, chum. Most people use stereo amplifers with independent L and R channels, and don't matrix music to M/S to distort it. I'm beginning to think that you are all charged up with the technology of making musical samples, and unthinkingly want to apply conventions from there to everything, and with attitude.


How is this remotely germane to this discussion?  You're trying to compare short-term loudness differences on peaks (dynamic range perception) to long-term loudness perception?  You've completely confused the entire issue, and confounded your test with something that just frankly seems, well, irrelevant.

I'm not doing studies in human perception or making samples, I'm simulating what real world audiophiles do.


If you add a little distortion, then it doesn't bring up the level of the "A" sample, and if you add a lot then the "A" sample sounds like it is harsher.

You seem to want to argue with someone, not really think about the nature of the question.

The claim has been made that equipment with modest amounts of distortion creates the perception of greater loudness than its actual power capabilities would suggest.  IOW a 50 watt tube amp can sound as loud as a 100 watt SS amp because of the additional nonlinear distortion of the tubed amp.


Since, then, we're talking about apparent dynamic range, and  in particular instantaneous loudness, you're not even testing the known facts. I have no idea, what, really, you are testing, but I have little doubt that if you raise the AVERAGE loudness by a factor of 1.21, give or take, by distortion mechanisms, it's going to sound bizzare.

What I find truely bizarre is the chip on your shoulder. You've obviously been studying Schoepenhauer, and applied his first strategem for winning any argument despite its merits. S's first strategem is to take your opponent's ideas beyond their natural range.


Of course, this has nothing whatsoever to do with the actual effects of loudness enhancement in a dynamic signal. So what have we proven here?

That you're not here to share ideas and opinions, but to make yourself look smart by resorting to intellectual trickery.


Conclusion: You may make something sound harsher, and some listeners may equate harsher with louder, but in the end the educated ear is not fooled.

Darn, that was what I said to start out with.


No, you have to provide affirmative evidence for your assertion, not negative evidence for something else.

Actually, I don't have to do anything in particular.


So you have no reasonable justification for that issue.

You seem to be confused about issues and observations. My observation was that I tried a certain approach to using nonlinear distortion to increase the effective loudness of some music without actually amplifying it signfiicantly. I found out that it wasn't that easy to do.

Since the claim is usually made by people who are randomly hooking up audio gear and arbitrarily choosing operational conditions, it seems like this isn't going to happen as desired very often.


Now, you have to figure out how to:

Compare perceived dynamic range of two signals via blind test of low-order distortions.
Find out how much dynamic range enhancement sounds "good" and how much sounds "bad".
Figure out how to avoid aliasing in your distortion mechanisms.
Figure out what distortions to use in M and S if you're trying to mimic vinyl.
Figure out what level of antiphase low-frequency distortion is relevant.
Figure out what level of low-level noise, and of what frequency, to use, if LP is what you're mimicing.

The job is not as simple as you appear to be claiming.

Not surprisingly, with all the smoke and fire you've supported my point. Plus, you've tipped your hand as someone who needs to be watched carefully because you want to win arguments, not actually share useful knowlege or personal experiences.

Moderation edit: cleaned the quotes up so that others and I can see more clearly what's going on.

Vinyl vs Digital and 24 bit vs 16 bit from vinyl.

Reply #99 –
Quote from: Arnold B. Krueger on
Quote from: Woodinville on

So, then, this is done in analog or digital domains?

Doesn't matter but I usually do stuff like this in the digital domain because its easier.

Actually, it matters a great deal how you did it.
Quote
Quote
If you did it in the digital domain, you've made an awful mistake and created a mass of aliased inharmonic components that are going to sound "very odd" at the least, unless you've done more than you indicated here. Did you oversample? Did you control your nonlinearities so that no products of the nonlinearity exceed fs/2?

I don't know who you think you are talking to, but you are obviously talking down/talking trash. But, I'll humor you with some straight answers to show good faith.

I know you, Arny, you know me.  I'd think the signature line would be sufficient, given the number of offensively rude things you've said about me in public fora...
Quote
(q) Did I oversample?

(a) Don't have to. I just used a sufficiently high sample rate that the signfiicant harmonics were in the usual operational range of the sample rate I used.
So, your oversampling rate was higher than the order of the nonlinearity? Yes or no. If you wish to argue this, then you have to explain precisely what sampling rate and what input bandwidth you had.  You have to have some level of oversampling, and you have to assure the input bandwidth or you will get in trouble.

If you're doing this for a stationary, sine-wave signal, you're excluding the subject of the test, so this has to be done with real, actual material, using a real, actual 20khz bandwidth.

So, then, if that IS what you did, then you had to have, with a 2nd order distortion function, at least a 40khz bandwidth, i.e. you had to sample at at least 88.2K, to use standard rates.

Is this what you did, yes or no? How did you ensure that no distortions (or combinations, sums, differences, all the results of distortion) did not alias?

Quote
I typically do work like this at 192/24 or higher, and downsample to 44/16 for delivery. I also carefully check my work with test signals to make sure that practical = theoretical. I've been known to use sample rates up to 10 MHz if my testing deems it important.
I see, then, you did oversample.  But as you say below, you don't know the effective order of the sharp clipping, which is very, very high order distortion.  That may or may not be a problem, though, but let's examine your statemetn that you didn't have to upsample.

You are exactly oversampling, distorting, and downsampling, so you could have been constructive and said "yes, I did" rather than saying "I didn't need to".  Your "sufficiently high sampling rate" is bandwidth limited at the input, so it's oversampled. Or, if it's not, you have effectively oversampled via your signal's actual bandwidth. So you did oversample. Of course, this begs the question of source, now, let me guess, was the source a CD original, right? So you had to get up to that higher sampling rate somehow, right?

Or did you take an analog signal into this higher sampling rate immediately. 

Unless you started with a completely analog music signal, you DID upsample one way or another, then you distorted, and then you downsampled. But you said you didn't need to. I can't really quite understand your picking a fight words here.
Quote
In this case I was working with added second hamonic distortion (an ideal SET) or sharp clipping (an amp that is clean within its ratings, but under rated).

Quote
What was the sample rate multiplier?

Since I know there is nothing magic about integer multiplers for upsampling, I used whatever it takes to get to 192/24.

Multiplier does not have to be integer.  Lots of people can multiply by n*48/44.1 or vice versa. Who used the word "integer", Arny? Why did you make that up completely?
Quote
Quote
Did you split it into M/S and use different distortions on M/S as is appropriate?

That isn't what happens in the real world, chum.


In case you don't know, and it appears you don't, LP distortion mechanisms are quite different for vertical vs. horizontal modulation of the groove. Vertical is L-R, i.e. S, and horizontal is L+R, i.e M.

Since these two have different distortion mechanisms, a reasonable way to take these issues into account is to work in a domain that is understood. This isn't a complete solution, of course, to a 2D nonlinear problem, but it's better than a 1D solution.

So, since you are mistaken at this point, I think you need to go back, study the LP distortion spectra for L-R and L+R (the old AES reprints book will give you the data you need), and then try that, in terms of comparing to the "loudest part" of music, rather than "average level".

Then, maybe, you'll have a meaningful experiment.  As it stands, the results of your experiment are not surprising, and tell us nothing about the effect on perceived dynamic range. They do confirm that lots of distortion is bad, which I think we all agree with.
-----
J. D. (jj) Johnston